Published September 1, 1987
Hexabrachion Proteins in Embryonic Chicken Tissues and Human Tumors
H a r o l d E Erickson a n d H o p e C. Taylor
Department of Anatomy, Duke University Medical Center, Durham, North Carolina 27710
Edelman, 1985, Proc. Natl. Acad. Sci. USA, 82:80758079) now appears to be identical to the chicken hexabrachion protein.
In a search for functional roles, we looked for a
possible cell attachment activity. A strong, fibronectinlike attachment activity was present in (NI'Ia)2SO4
precipitates of cell supernatant and sedimented with
hexabrachions in glycerol gradients. Hexabrachions
purified by antibody adsorption, however, had lost this
activity, suggesting that it was due to a separate factor
associated with hexabrachions in the gradient fractions. The combined information in the several, previously unrelated studies suggests that hexabrachions
may play a role in organizing localized regions of extracellular matrix. The protein is prominently expressed at specific times and locations during embryonic development, is retained in certain adult
tissues, and is reexpressed in a variety of tumors.
s a previous paper (11) we described the structure of
an extracellular matrix protein produced by chicken
embryo and human fibroblasts. The protein was a
disulfide-bonded oligomer of 220 kD (chicken) or 285 kD
(human) subunits. Electron microscopy of both the chicken
and the human oligomers revealed a distinctive six-armed
structure that we called a hexabrachion. These oligomers had
previously been considered a form of fibronectin (1). There
were several distinctive structural features, however, that
suggested this was a different protein.
Two proteins extensively characterized by other laboratories seemed very similar to the hexabrachion proteins. Chiquet and Fambrough (7, 8) described a protein from chicken
tissues and cell cultures that they called myotendinous (M1)
antigen. This protein was a disulfide-bonded oligomer of
220-kD subunits, with variable components of lower molecular mass. It had the same sedimentation coefficient as our
chicken hexabrachions. Bourdon et al. (2-4) described a
protein called glioma-mesenchymal extracellular matrix
(GMEM) 1 protein that was prominently expressed in gliomas and some other human tumors. The antigen was not
detected in normal brain tissue but was present at specific lo-
cations in several other tissues. This protein was a disulfidebonded oligomer of 250-kD subunits, similar to the hexabrachion protein from human fibroblasts. For the present
study we obtained the M1 and 81C6 monoclonal antibodies,
which originally identified the myotendinous antigen and
GMEM protein, and used them to purify the oligomers from
chicken and human cell cultures. Electron microscopy
confirmed that both these antigens are hexabrachions, and
we will argue that they must be homologous.
More recently a protein called cytotactin has been identiffed in embryonic chicken brain and several other chicken
tissues (10, 15). We noted that the distinctive band pattern of
cytotactin on gel electrophoresis was identical to that of the
Mt antigen. The tissue distributions were also quite similar,
with one important exception: whereas the previous study of
M1 antigen reported no detectable protein in brain, embryonic chicken brain was the primary source for cytotactin.
Here we address this apparent discrepancy and show that M1
antigen is actually abundant in embryonic chicken brain.
In our previous study we identified a strong hemagglutination activity associated with the hexabrachions (11). Because
hemagglutination is frequently associated with cell attachment, we wanted to determine directly whether the hexabrachions could mediate cell attachment. While preliminary
results demonstrated a strong cell attachment activity very
similar to fibronectin (12), it now appears that this activity
I
L Abbre~ations used in thispaper: CAPS, 2-(cyclohexylamino)-l-propanesulfonie acid; GMEM, glioma-mesenchymal extracellular matrix; pFN,
plasma fibronectin.
9 The Rockefeller University Press, 0021-9525/87/09/1387/8 $2.00
The Journal of Cell Biology, Volume 105, September 1987 1387-1394
1387
Downloaded from jcb.rupress.org on February 21, 2013
Abstract. Cell cultures of chicken embryo and human
fibroblasts produce a large extracellular matrix molecule with a six-armed structure that we called a hexabrachion (Erickson, H. P., and J. L. Iglesias, 1984,
Nature (Lond.), 311:267-269. In the present work we
have determined that the myotendinous (M1) antigen
described by M. Chiquet and D. M. Fambrough in
chicken tissues (1984, J. Cell Biol., 98:1926-1936),
and the glioma mesenchymal extracellular matrix protein described by Bourdon et al. in human tumors
(Bourdon, M. A., C. J. Wikstrand, H. Furthmayr,
T. J. Matthews, and D. D. Bigner, 1983, Cancer Res.
43:2796-2805) have the structure of hexabrachions. We
also demonstrate that the M1 antigen is present in embryonic brain, where it was previously reported absent, and have purified hexabrachions from brain
homogenates. The recently described cytotactin (Grumet, M., S. Hoffman, K. L. Crossin, and G. M.
Published September 1, 1987
is associated with but not an integral part of the hexabrachions.
Brachionectin has been proposed as a general name for the
group of proteins with the hexabrachion structure (12, 19).
Tenascin has been proposed as a new name for myotendinous
antigen of chicken and immunologically related proteins
from other species (9). Here we will avoid controversy over
nomenclature and refer to the proteins as hexabrachions or
hexabrachion proteins, a usage that will emphasize the
remarkable structural homology.
Materials and Methods
Cell Lines and Culture Techniques
Purification of Hexabrachions from Cell Supernatant
Cells were grown to confluence in 150-cm2 flasks (Coming Glass Works,
Coming, NY) with 25 mi of medium per flask. Typically five flasks were
used for each preparation. Cell supernatant was removed for processing on
the day of confluence and at 2-3 d intervals afterwards. At each of these
times fresh medium was added. Protein was harvested from fibroblast cultures until 10-12 d after confluence, and from the U-251 MG glioma cells
until 18 d.
Immunoaffinity columns were prepared by coupling 5 mg of M1 or 81C6
antibody to 1 g (3.5 ml) of cyanogen bromide-activated Sepharose. Typically 150 ml of cell supematant was centrifuged at 20,000 rpm for 20 min
to remove cellular debris, and then passed over the column. The column
was washed with 50 ml of 0.1 M NaBO4, 0.5 M NaCI, pH 8.4 and then
eluted with 100 mM CAPS (2[cyclohexylamino]-l-propenesulfonic acid),
0.5 M NaCt, pH 11 as described by Chiquet and Fambrough (7). Immunoaflinity chromatography was done at room temperature.
Hexabrachions were alternatively isolated from cell superuatants by
precipitation with (NH4):SO4 and gradient sedimentation. 20 g (NI-hhSO4
was added per 100 ml initial volume, giving a 33 % saturated solution. The
pH was adjusted to ,x,7.4 with NI-I4OH, the material stirred for 30 min at
room temperature and centrifuged 30 min at 30,000 rpm in a Beckman-type
35 rotor (Beckman Instruments, Inc., Palo Alto, CA). The pellets were
resuspended in 5 ml of 200 mM CAPS, 0.15 M NaCI, pH 11. The protein
was precipitated a second time by adding (NI~hSO4 (with no pH adjustment) and the pellet resuspended in 0.6-1.5 ml CAPS buffer.
Gradient sedimentation was performed by layering a 0.5-mi sample of
protein in CAPS buffer on a 15--40% (vol/vol) glycerol gradient containing
0.2 M NI-hHCO3, pH 9.5. The gradients were centrifuged at 41,000 rpm in
a Beckman SW41Ti rotor, 20~ for 18 h. 20 fractions were collected from
a hole in the bottom of the tube. Sedimentation coefficients were estimated
from standards (fibrinogen at 8 S in fraction 12, catalase at 11.3 S in fraction
9, and alpha-2 macroglobulin at 18.5 S in fractions 3 and 4) run in a separate
gradient.
Protein Assay and Gel Electrophoresis
Protein concentrations were determined by the method of Bradford (5),
using reagents from Bio-Rad Laboratories (Richmond, CA). Values were
referred to a standard curve of highly purified human plasma flbronectin
(pFN). The concentration of this pFN was established by UV spectroscopy,
using an extinction coefficient of 1.28 cm2/(mg/ml).
PAGE was performed according to Laemmli (18), using a 5 % running
gel and a 3 % stacking gel. Samples were dialyzed into CAPS buffer and
mixed 1:1 with sample buffer containing 1% SDS and 1% beta-mercaptoethanol. Electrophoresis was for •4 h at 200 v, to bring the dye front to
the bottom of the 13.5-cm gel. The gel was washed overnight in 50% methanol, 12% acetic acid and stained with silver by the method of Merril et al.
(20).
Immunoblotting was performed after transfer to nitrocellulose using a
semi-dry apparatus as described by Gibson (14). Protease-faeilitated transfer
(14) gave the most efficient transfer of the ~600-kD band, but it was only
useful when blotting with the polyclonal antibody. The epitopes for the
monoclonal antibodies were destroyed by the pronase.
Electron Microscopy
10-~tl samples of protein in 0.2 M NI-LHCO3, 30% glycerol, pH 9.5 were
sprayed onto mica, dried in vacuum, and rotary shadowed (13). Mierographs
were taken at 50,000x and printed at a final magnification of 150,000x.
Immunocytochemistry
An ll-d chicken embryo brain was fixed with 2 % paraformaldehyde and
glutaraldehyde. 50-1am sections were cut with a vibratome, stained with M1
or control antibody at 10 ~tg/ml for 18 h at 4~ and then with peroxidasecoupled anti-mouse IgG (Boehringer Mannheim Diagnostics, Inc., Houston, TX). The peroxidase was developed with diaminobenzidine, the section was postfixed with osmium tetroxide and embedded in Embed-812 (EM
Sciences, Cherry Hill, NJ).
Cryostat sections (6 I~m) of frozen, unfixed chicken brain were stained
with M1 antibody and fluorescein-conjugated second antibody.
Cell Attachment Assay
Cell attachment was assayed by the microtiter well method described by
Ruoslahti et al. (22). Serial dilutions of fibronectin or the protein fractions
to be tested were incubated in the plastic wells for 2 h at room temperature,
the wells were blocked with 1% heat (g0~ 3 rain) denatured BSA and
washed. Cells were left on the substrate at 37~ for 1 h. After washing off
unattached cells and staining, the assay was scored by visual inspection of
the wells in a low power microscope, looking for the minimnn protein concentration that would give significant attachment.
Results
PurificationofHexabrachionsby Immunoadsorption
Brain, gizzard, or wings from ll-d chicken embryos were homogenized in
200 mM CAPS, 0.15 M NaC1, pH 11 (5 ml buffer per gram of tissue) for
1 min in a Sorvalt Div. Onmimixer (Newton, CT). The homogenate was left
on ice for 30-60 min, and then neutralized to pH 7-7.5 by adding t M
NaH2PO4. Phenylmathylsulfonyl fluoride was added to a concentration of 1
raM. Brain homogenates were processed with no further additions, but giz-
Fig. 1 shows analysis by PAGE of the proteins purified from
cell supernatants and tissue homogenates using the MI antibody for chicken and the 81C6 antibody for human material.
Human and mouse hexabrachions purified by (NH4)2SO4
precipitation and gradient sedimentation, without antibody
adsorption, are also shown in Fig. 1 and discussed in a later
section.
The M1 antigen purified from chicken fibroblast cell supe~atant by antibody adsorption (M1-CEF) comprises a
prominent band at 220 kD and a closely spaced doublet near
200 kD. This is the pattern previously described by Chiquet
and Fambrough (8) for the M1 (myotendinous) antigen, and
The Journal of Cell Biology, Volume 105, 1987
1388
Preparation of Hexabrachion Protein from
1issue Extracts
Downloaded from jcb.rupress.org on February 21, 2013
Primary cultures of chicken embryo fibroblasts were prepared from skin of
10-d embryos. Cultures were used for protein preparation from the fourth
passage until they stopped growing, usually tenth passage. U-251 MG, clone
3, an established line of human glioma (astrocytes), were obtained from Dr.
D. D. Bigner, Duke University. Mouse fibroblast 313 cells were obtained
from the Lineberger Cancer Center, University of North Carolina. NIL.8M
hamster fibroblasts were obtained from Dr. Richard Hynes, Massachusetts
Institute of Technology.
Cells were grown in DME supplemented with 10% FCS in a 5% CO2
atmosphere. After the cells reached confluence they were fed with serum
that was pretreated with 33% (saturated) (NFI4hSO4 to remove serum proteins that would precipitate.
zard and wing homogenates were usually treated with DNAase and
hyaluronidase to reduce viscosity. The homogenate was centrifuged at
35,000 rpm for 45 rain to remove particulates, and the superuatant was
filtered through Whatman No. 1 paper (Whatman Inc., Clifton, NJ) to remove floating debris. The solution was then passed over the M1 antibody
column, the column was washed, and the protein eluted as described above.
Published September 1, 1987
blasts (duplicate samples shown in lane 2 and far right). The next
three lanes show M1 antigen purified from tissue homogenates of
ll-d chicken embryos. Ml-brn, brain; M/-gzz, gizzard; M/-wng,
wing (the brain and wing lanes are partly covered to show heavily
stained bands). 81C6-gl, 81C6 antigen purified from cell supernatants of U-251 MG human glioma cultures. Gr6-gl and Gr7-gl,
fractions 6 and 7 from a glycerol gradient of (NI-I4)2SO4precipitate of glioma cell supernatant. GrT-3T3,fraction 7 from (NH4)2SO4 precipitate of cell supernatant from mouse 313 cells. Two different preparations of Gr7-3T3 are shown. Subunit Mr x 10-3,
shown on the right, were estimated from the following standards:
440, laminin, upper band; 220, pFN, upper band of doublet; 200,
myosin; 145, muscle C protein.
also the pattern reported by Grumet et al. (15) for cytotactin.
Both of these reports assigned a value of 190-200 kD to this
doublet so we will continue to use these numbers, although
on our gels the doublet actually ran slightly above myosin
(usually assigned a value of 200 kD as an Mr standard). The
intensity of the doublet relative to the 220-kD band was variable, as previously noted by Chiquet and Fambrough (8).
The M1 antigen was also isolated from homogenates of
several chicken tissues, all from ll-d embryos. Brain homogenates gave the same three bands as the fibroblast cell supernatant, although with different relative intensities, plus new
bands. The 190-kD band, the lower band of the doublet, was
strong in the brain protein. A cluster of bands of higher mass,
~500 kD, was apparent in some preparations. M1 antigen
was also obtained from gizzard and wing homogenates. The
190-kD band was prominent, the other bands of brain extracts were present, and there were additional bands at
higher and lower mass. The heterogeneity of subunit sizes is
particularly evident in the antigen from wing homogenates
(Ml-Wng).
The 81C6 antigen from U-251 MG glioma cells was previously identified only in the extraceUular matrix, and was
thought to be absent from the cell supernatant (4). In our
present study we found it to be abundant in the cell superna-
Ericksonand TaylorHexabrachionProteins
Purification of Hexabrachions from Cell
Supernatants without Antibody Adsorption
We had previously purified hexabrachions from cell supernarants using (NI-hhSO4 precipitation followed by gradient
sedimentation (11). This procedure was especially useful
with human hexabrachions, which sedimented faster than
those from chicken and were well separated from fibronectin. Hexabrachions from human U-251 MG glioma cells
sedimented at 14-17 S. A 15-S fraction (lane Gr6-gl in Fig.
1) showed a prominent band at 285 kD, a weak band at ~600
kD, and two bands near 200 kD. In the next fraction (13.5 S,
lane Gr7-gl) only this doublet is prominent, Fraction 6
showed a high concentration of hexabrachions by electron
microscopy; fraction 7 had far fewer hexabrachions. Thus
the strong band at 285 kD corresponds to the presence of
hexabrachions, while the doublet at 200 kD (which is much
stronger and also more widely spaced than the doublet in
antibody-adsorbed protein) may be contaminating proteins.
Figure2. Immunoblots stained
with the 81C6 monoclonal
against GMEM (a and b); M1
monoclonal against chicken
myotendinous antigen (c and
d); and Chiquet-Fambrough
polyclonal against M1 antigen
(e-i). The proteins loaded
are: (a) gradient purified human hexabrachions, a fraction
similar to Gffa-gl (Fig. I) but
with a heavier -.~a00-kDband;
(b) human hexabrachionspurified by 81C6 antibody absorption, similar to 81C6-gl (Fig.
1) but with a lighter ~600-kD
band; (c) chicken hexabrachions purified from fibroblast cultures by M1 antibody adsorption;
(d) chicken hexabrachions purified from brain homogenates by M1
antibody adsorption; (e) human fibronectin, 2 x more protein than
lanesf - i ; (f) gradient purified human hexabrachions; (g) antibody
purified human hexabrachions; (h and i) two concentrations of
chicken embryo fibroblast M1 antigen.
1389
Downloaded from jcb.rupress.org on February 21, 2013
Figure 1. Gel electrophoresis of hexabrachion proteins purified by
antibody adsorption (lanes M/and 81C6) or by glycerol gradient
sedimentation (lanes Gr6, 7). The lanes are given descriptive labels. pFN, human plasma fibronectin (0.5 Ixg protein loaded); M/CEF, M1 antigen from cell supernatant of chicken embryo fibro-
tant. Protein isolated from cell supernatant by adsorption to
the 81C6 antibody column showed a prominent and somewhat diffuse band (Fig. 1, 81C6-gl) significantly higher than
the 220-kD band of the chicken antigen. We previously estimated a mass of ,'o285 kD for this band from human fibroblasts (11), while Bourdon et al. (4) and Carter and Hakomori
(6) both designated it 250 kD. The band ran about half-way
between reduced and nonreduced plasma fibronectin (220
and 440 kD), so the higher value seems more appropriate.
The 81C6 antigen also showed a closely spaced doublet that
ran slightly above the chicken doublet. This doublet was always very weak in the human protein. Finally, a sharp band
of much higher mass, running above nonreduced pFN and
the upper band of laminin, was apparent in most preparations. We estimate a value of * 6 0 0 kD for this band, but
there are no standards in this region. Neither this high molecular mass band nor the doublet was observed in the previous study of GMEM protein (4). The sensitive silver stain
and the larger quantities of protein obtained by our methods
were important to demonstrate these weaker bands.
Published September 1, 1987
Cell supernatant from mouse 313 cells was processed by
(NH4)2SO4 precipitation and gradient sedimentation. The
13.5-S fractions showed a set of bands near 200 kD (Fig. 1,
lanes Gr7-3T3)and electron microscopy showed hexabrachions (Fig. 4 E). This purification scheme, although not as rigorous as antibody adsorption, is quite useful for screening
cell cultures for production of hexabrachions.
Another advantage of the gradient purification is the high
yield of protein. The peak hexabrachion fractions from
glioma cultures had a protein concentration up to 0.5 mg/ml
(from 200 ml U-251/rig cell supernatant), in which the 285
kD appeared to account for >80% of the band intensity on
silver-stained gels. In contrast the 5-ml antibody columns
appeared to be saturated at ,o200 gg of protein, and the maximum concentration of eluted protein was "o50 ttg/ml.
Immunological Cross-Reactivity and
lmmunocytochemistry
Figure3. (A) A horizontal section through the optic tectum of an ll-d chicken embryo stained with the M1 monoclonal antibody and visualized with peroxidase-conjugated second antibody. The antibody stain is dark. Staining is pronounced in two layers adjacent to the neuroepithelium (the letter v indicates the ventricle). Staining of the external layers is much lighter, but darker than the control. The dark
lines are capillaries (c) which stain because of endogenous peroxidase activity of blood cells. (B) A frozen section of brain tissue stained
with M1 antibody and a fluorescein-conjugated second antibody. The (white) fluorescence is intense in punctate patches between cell bodies
on the right, but is virtually absent from the ,,o20-~tmsurface layer at the bottom. The arrow indicates the ventral edge of the brain tissue.
Bar in A, 100 ~tm; in B, 10 txm.
The Journal of Cell Biology.Volume 105, 1987
1390
Downloaded from jcb.rupress.org on February 21, 2013
Immunoblots of human and chicken hexabrachion proteins
are shown in Fig. 2. The 81C6 antibody stained both the main
285-kD band and the "o600 kD of human hexabrachion protein (Fig. 2, a and b). A diffuse patch from 260-220 kD also
stained, but the pronounced doublet in gradient fraction Gr6gl did not stain. The MI antibody stained the main 220-kD
band and the 190-200-kD doublet of M1 antigen from fibroblast cell cultures (c), as previously demonstrated by Chiquet
and Fambrough (8). These same bands, as well as a weak
210-kD band, were stained in M1 antigen prepared from
brain homogenate (d). The "o500-kD bands seen in Fig. 1,
Ml-brn, were missing from this preparation, but when present they also stained with the M1 antibody. Neither the 81C6
nor the M1 monoclonal antibodies cross-reacted with other
species.
Chiquet and Fambrough prepared a polyclonal antibody to
the chicken M1 antigen and showed that it did not cross-react
with fibronectin. In Fig. 2, h and i we show that this polyclonal antibody stains the main band and the doublet of M1
antigen from chicken cultures, as well as a "o500-kD band
and several lower Mr bands, which were scarcely visible on
the silver-stained gel of this preparation. Most important,
this polyclonal antibody cross-reacts with the human hexabrachion protein, staining both the 285- and the "o600-kD
bands.
The presence of M1 antigen in embryonic chicken brain
was confirmed by immunocytochemistry. Fig. 3 A shows a
Published September 1, 1987
horizontal section through the optic tectum of an ll-d chicken
embryo. This thick section, stained with the M1 antibody and
a peroxidase-conjugated second antibody, shows a concentration of M1 antigen in certain layers of the tectum, probably
layers II and HI as defined by LaVail and Cowan (17), with
much lighter staining of other layers. Fig. 3 B shows a thin
section of brain tissue stained with a fluorescent second antibody. The M1 antibody did not stain the outermost (neuroepithelial) layer but was localized in patches around the
cell bodies in the deeper layers.
Electron Microscopy of Purified Hexabrachions
Erickson and Taylor Hexabrachion Proteins
Figure 4. Electron micrographs of hexabrachions from human,
chicken and mouse. Bar, 100 ~tm. 10-~tl samples of glycerol gradient fractions were sprayed onto mica and rotary shadowed (11).
(A) Hexabrachions from human glioma cells, purified by (NH4hSO4 precipitation and gradient sedimentation (Fig. 1, lane Gr6gl). Human hexabrachions purified by antibody adsorption were
identical in structure. (B) Hexabrachions from chicken embryo
fibroblasts, purified by immunoafiinity followed by gradient centrifugation (13-S peak). (C) Trimers (half-hexabrachions) from
chicken embryo fibroblasts (9 S peak). (D) M1 antibody (10 txg/ml)
reacted with chicken embryo fibroblast hexabrachions (concentration not determined, ~10 ~tg/ml) in gradient buffer, overnight at
4~ Antibody molecules are bound to the arms about one-third of
the distance from the center. (E) A hexabrachion from 3T3 mouse
fibroblasts (Fig. 1, lane GrT-3T3).
of hexabrachions are sometimes found to be attached to each
other, as in the left hand panels in Fig. 4, A and B. These
complexes are not frequent, but the morphology is very reproducible: a reciprocal contact of a distal knob of one hexabrachion at a midpoint of an arm on the other. These images
suggest the possibility that hexabrachions can self-associate
to form pairs and eventually more extended networks.
1391
Downloaded from jcb.rupress.org on February 21, 2013
Proteins purified from chicken embryo fibroblasts or human
glioma cells by antibody adsorption and/or by gradient sedimentation were examined by rotary shadowing electron
microscopy (Fig. 4). Protein eluted from the antibody columns always showed a high concentration of hexabrachions,
but these preparations were frequently contaminated with
irregular debris, perhaps shed from the column. Fractions
from a subsequent gradient sedimentation showed clean
hexabrachions, their concentration corresponding to the intensity of the bands on gels. Thus the M1 and the 81C6 antigens are identified as chicken and human hexabrachion proteins, respectively.
In addition to typical hexabrachions, the chicken protein
showed a significant number of three-armed oligomers. These
trimers were separated from hexabrachions by gradient sedimentation. Hexamers were obtained in a 13-S peak, and
trimers and some single strands were found in a 9-S peak.
The trimers (Fig. 4 C) are clearly half-hexabrachions: the
arms are arranged in the "T" configuration, usually with a
central nodule (or part of the central nodule) attached above
the cross of the "T" Each arm has the knob on the end, and
has the characteristic thickening of the distal segment. We
can't yet say whether these half-hexabrachions are products
of proteolysis or of incomplete assembly.
The binding of the M1 antibody to chicken hexabrachions
was visualized directly in electron microscope specimens. In
this experiment hexabrachions were prepared by (NI-h)2SO4
precipitation and gradient sedimentation, without exposure
to the antibody column. The peak fraction of hexabrachions
as assayed by electron microscopy was incubated with M1
antibody in the gradient buffer. An antibody concentration
of 10-30 ~tg/ml gave the cleanest specimens (the concentration of hexabrachions was estimated to be '~10 txg/mi). After
overnight incubation at 4~ and 1 h at room temperature,
samples were sprayed for rotary shadowing. Antibody molecules could be seen clustered near the center of the hexabrachion, apparently binding to the artns about one-third of
the distance from the center (Fig. 4 D). Similar attempts to
localize the 81C6 antibody on human hexabrachions have not
been successful, perhaps because of low affinity.
A mouse hexabrachion from 3T3 cell supernatant, isolated
by (NI-L)2SO4precipitation and gradient sedimentation (Fig.
1, lane Gr7-3T3), is shown in Fig. 4 E.
Two points of structure should be noted in the examples
shown in Fig. 4. First, the arms are usually curved or sharply
kinked, and in almost all cases this bend is in the clockwise
direction. This means that the hexabrachions have a preferred orientation when they flatten on the mica, probably
because they are cup shaped rather than flat. Second, pairs
Published September 1, 1987
Cell Attachment Activity
Figure5. A diagram of a hu-
man hexabrachion (reprinted
from reference 11 with permission) designating five characteristic features. Each of the
arms has (a) a terminal knob,
(b) a thicker distal segment,
and (c) a thinner inner segment. Three arms come together to form a "T" junction
(d) at which two arms extend
colinearly in opposite directions and the third is perpendicular. Each of these trimers
is attached by a short link to a central globular particle (e) As shown
in Fig. 4 C the trimers can be isolated as separate entities. Bar,
100m.
~b
The myotendinous (M1) antigen from chicken (7, 8), and the
GMEM protein from human tumors (2-4) have each been
well characterized biochemically and histologically, but not
previously associated with each other. The evidence we present here shows that these two proteins are structurally and
immunologically homologous. Correlating these previously
unrelated studies, and the more recent studies of cytotactin,
suggests that hexabrachion protein is prominently expressed
in specific locations during embryonic development, includ-
ing development of the brain, and is reexpressed by several
types of tumors.
The strongest argument for homology of the chicken embryo and human tumor proteins is their distinctive and elaborate hexabrachion structure. With most protein molecules a
low resolution structure would not be sufficient basis for suggesting homology. The hexabrachion, however, has at least
five distinctive features, illustrated in Fig. 5. It is difficult to
imagine that all the features of this complex structure could
arise in unrelated proteins.
The homology is confirmed by immunological cross-reactivity. The monoclonal antibodies stain all the major bands
of their own antigen (8; and Fig. 2), but are not cross-reactive with other species. A polyclonal antibody against the
chicken M1 protein stained the main 285-kD and the ,0600kD bands of human hexabrachions.
A curious feature of the human protein is that its major
subunit is much larger than that of chicken, 285 kD vs. 220
kD. In our previous study (11) we showed that this size difference corresponded to the length of the arms: the arms of
chicken hexabrachions were 68-urn long and those of human,
87-urn long. The smaller subunits of the chicken doublet
might correspond to still shorter arms but we have not yet
been able to demonstrate this small difference.
The multiple bands seen especially with the chicken protein may be the result of alternative RNA splicing. The bands
differ in Mr by 5-20 kD, and the larger ones all stain with
the monoclonal antibodies. Data of Chiquet and Fambrough
(8) suggest that the heterogeneity is not due to proteolysis.
Fibronectin, which has a strand-like structure similar to the
hexabrachion arm, is a string of small (5-10 kD) globular domains, some of which are omitted by alternative RNA splicing (21). An analogous mechanism seems an attractive hypothesis to explain the heterogeneity of hexabrachion bands.
We can now make a strong argument that the recently
named cytotactin is the same as the M1 antigen, the chicken
hexabrachion protein. The band pattern on gel electrophoresis reported for cytotactin (15), both reduced and nonreduced, is identical to that reported here and previously (8)
for the M1 antigen. Our present finding that M1 antigen is
prominent in embryonic chicken brain resolves the only apparent discrepancy between the two proteins. The identification of hexabrachions in embryonic brain adds a new dimension of interest to the protein-developmental neurobiology.
The detailed studies of cytotactin have already demonstrated
The Journal of Cell Biology, Volume 105, 1987
1392
Hexabrachion Synthesis by Different Cell Lines
The U-251 MG human glioma cultures were the richest cellular source of hexabrachions we have found, 200 ml of cell
supernatant yielding up to 1 mg of protein in the hexabrachion gradient fractions. Several other cell types were assayed for hexabrachion production by preparing gradient
fractions from cell supernatant and looking for hexabrachions by electron microscopy. Primary cultures of chicken
embryo and human fibroblasts produced much less hexabrachion protein and relatively more fibronectin than the U-251
MG. Established fibroblast lines, 313 from mouse and normal rat kidney from rat, also produced hexabrachions. Cultures of virally transformed chicken embryo fibroblasts, 3T3
and normal rat kidney cells all produced hexabrachions, essentially the same as the parent cultures. This establishes an
important point, that hexabrachion production does not seem
to be linked to viral transformation.
Discussion
Downloaded from jcb.rupress.org on February 21, 2013
Using the microtiter cell attachment assay of Ruoslahti and
Pierschbacher (23) we obtained a reproducible titration of
cell attachment activity with pFN. Wells coated at 5-10
ltg/ml pFN were mostly covered with flattened cells, and
substantial cell attachment was obtained down to 2.5 gg/ml
pFN. NIL.8M hamster fibroblasts and U-251 MG glioma
cells gave virtually identical results in attaching to pFN.
Human hexabrachions prepared from cell supernatants by
(NH4)2SO4 precipitation and gradient sedimentation reproducibly showed a high level of cell attachment activity (12).
The hexabrachion fractions promoted cell attachment down
to a protein concentration of 3.5 gg/ml, and the attachment
was inhibited by the peptide GRGDSP, which is known to inhibit cell attachment to fibronectin and other proteins (23).
The 285-kD hexabrachion band accounted for ,080% of the
intensity seen on silver-stained gels (Fig. 1, lane GR6-gl).
Bands ,0200 kD, not seen on antibody purified hexabrachions, are the most obvious containment.
Hexabrachion protein purified by immunoaffmity chromatography, however, gave no cell attachment up to the highest concentrations tested, 20-50 gg/ml. Both NIL.8M and
U-251 MG cells, and hexabrachions from chicken and human, were tested. To address the possibility that the cell attachment activity was being destroyed by the pH 11 elution
buffer, separate experiments were performed in which hexabrachions were eluted from the antibody column with 2 M
urea or with 2 M potassium thiocyanate. These preparations
also had no cell attachment activity. We conclude that purified hexabrachions have no cell attachment activity, but an
active factor is associated with them in the gradient fractions.
/8
Published September 1, 1987
Erickson and Taylor Hexabrachion Proteins
induced breast tumors. Their study thus demonstrated the
same association with embryonic and cancerous tissue that
we have argued here.
A central question is the function of this extracellular matrix protein. The developmental regulation of its appearance
in embryos (7, 10), and its apparent reexpression in a variety
of tumors (2, 9) are especially intriguing. One would hope
that a comprehensive catalogue of its highly restricted tissue
distribution (2, 7, 9, 10, 19) would correlate with an obvious
function, but this is not the case. As an extracellular matrix
molecule it is reasonable to think that it must be binding to
cell surface receptors and/or other matrix molecules. The
highly specific tissue localization, e.g., in tendons, perichondrium, and perineureum of chicken embryos (7, 10), in
a very narrow sheath around bird muscle spindle (19), and
around capillaries and extracellular matrix fibers in human
tumors (2) suggest that hexabrachions may be binding to
other matrix molecules and possibly self-associating to establish the restricted distribution.
The structure of the hexabrachion suggests three points
that may be important in relation to function. First, it is big.
Each arm is 87-nm long (for the human; 68 nm for the
chicken, reference 11), so the arms can easily span a distance
of 150 nm. Second, the arms have a thin elongated structure.
The similar structure in fibronectin is thought to be based on
a linear arrangement of multiple small domains (21). If the
hexabrachion arms are also a linear array of small domains,
each arm could contain multiple independent binding functions. Third, the molecule is apparently multivalent. Assuming each arm is identical, the hexabrachion should be able
to bind six ligands into a single complex, or self-associate
with up to six other hexabrachions. This might easily lead
to formation of an extended network. Whatever functions are
eventually determined, it would be surprising if they did not
use some or all of the possibilities provided by the elaborate
hexabrachion structure.
We thank Douglas Fambrough, Johns Hopkins University, for the M1 and
polyclonal antibodies to myotendinous antigen. We thank Daretl D. Bigner,
Duke University, for the U-251 MG cells and 81C6 antibody to the GMEM
protein. Gina Briscoe provided excellent technical assistance with cell culture, attachment assays, and protein purification. Donna Bunch-O'Dell
provided assistance and advice on culturing different cell types. Francine
Gumkowski prepared the specimens for immunocytochemistry. E. T.
O'Brien performed early electron microscopy localizing the M1 antibody
site.
This work was supported by National Institutes of Health grants R01HL23454 and 5P30-CA14236.
Received for publication 14 October 1986, and in revised form 24 February
1987.
References
I. Alexander, S. S., G. Colonna, K. M. Yamada, I. Pastan, and H. Edelhoch.
1978. Molecular properties of a major cell surface protein from chick embryo fibroblasts. J. Biol. Chem. 253:5820--5824.
2. Bourdon, M. A., C. J. Wikstrand, H. Furthmayr, T. J. Matthews, and
D. D. Bigner. 1983. Human glioma-mesenchymal extracellular matrix
antigen defined by monoclonal antibody. Cancer Res. 43:2796-2805.
3. Bourdon, M. A., R. E. Coleman, R. G. Blasberg, D. R. Groothuis, and
D. D. Bigner. 1984. Monoclonal antibody localization in subcutaneous
and intracranial human glioma xenografts: paired-label and imaging analysis. Anticancer Res. 4:133-140.
4. Bourdon, M. A., T. J. Matthews, S. V. Pizzo, and D. D. Bigner. 1985.
Immunochemical and biochemical characterization of a glioma-associated extracellular matrix glycoprotein. J. Cell Biochem. 28:183-195.
5. Bradford, M. 1976. A rapid and sensitive method for the quantitation of
1393
Downloaded from jcb.rupress.org on February 21, 2013
a changing pattern of expression at specific times and locations during neural development, possibly related to cell
migration (10, 15).
Bourdon et al. (2) reported that the human GMEM protein
was prominent in most gliomas but absent from normal brain
tissue. We have not yet reinvestigated the presence of hexabrachion protein in human brain, but we would expect it to
be present, at least in fetal tissue.
The J1 antigen from mouse brain (16, 24) has been less extensively characterized than the other antigens, but it may be
the mouse hexabrachion protein. Some similarities to cytotactin have been noted (15). The GP-250 protein, previously
characterized as a component of the extracellular matrix of
human fibroblasts (6), is almost certainly the same protein
as the human hexabrachion. An important finding from this
study, also supported by experiments of Bourdon et al. (4),
is that hexabrachions can be extracted from the matrix by detergent in the absence of reducing agent. This is in contrast
to fibronectin, most of which is covaiently bound into the
matrix by disulfide bonds.
Our present finding that hexabrachions can be extracted intact from homogenates of chicken tissue makes two important points. First, it confirms that the association of hexabrachions with cells or other matrix components in tissues
is noncovalent, as it appears to be in cell culture. Second,
it shows that at least a significant fraction of the protein exists
in tissue with the intact hexabrachion structure, i.e., it is not
degraded or processed to other forms.
Hexabrachion protein purified by antibody adsorption had
no cell attachment activity comparable to fibronectin. There
is, however, a strong, fibronectin-like cell attachment activity associated with hexabrachions in glycerol gradient fractions of cell supernatants (12). The simplest explanation is
that this strong attachment activity is due to a separate factor,
associated with hexabrachions in the gradient fractions but
washed off during antibody adsorption. We should note that
the assay used in these studies (22) measures ceils that are
strongly attached by a combination of initial attachment and
spreading over the 1-h incubation. We are currently investigating the possibility that hexabrachions have a weak or restricted cell attachment activity not demonstrated by this particular assay.
Grumet et al. (15) reported that the cytotactin antibody
inhibited neuron-glial attachment, but in the experiment
shown the fraction of cells affected was very small, ,,o1% of
the total probe cells. Kruse et al. reported a much larger
effect with their J1 antibody (16) but details of their assay have
not yet been published. Both these studies report a limited
number of assays (two or three) and are using a complex
mixed cell assay. We are hesitant to ascribe a cell attachment
activity until it can be demonstrated with purified protein
using a solid phase assay.
In a recent paper Chiquet-Ehrismann et al. (9) found no
detectable cell attachment activity for the M1 "antigen, which
they renamed tenascin, using an assay involving 2 d of growth
on a protein-coated substrate. They did, however, observe
that the protein stimulated growth and cell division in the absence of added serum, a finding of potential significance to
the role of the protein in tumors. They identified an immunologically cross-reactive protein in rat and found that this protein was present in embryonic breast tissue, disappeared in
adult, and was prominently reexpressed in experimentally
Published September 1, 1987
6.
7.
8.
9.
10.
ll.
12.
13.
14.
15.
microgram quantities of protein utilizing the principle of protein-dye
binding. Anal. Biochem. 72:248-254.
Carter, W. G., and S. Hakomori. 1981. A new cell surface, detergentinsoluble glycoprotein matrix of human and hamster fibroblasts. J. Biol.
Chem. 256(13):6953-6960.
Chiquet, M., and D. M. Farnbrough. 1984. Chick myotendinous antigen.
I. A monoclonal antibody as a marker for tendon and muscle morphogenesis. J. Cell Biol. 98:1926-1936.
Chiquet, M., and D. M. Fambrough. 1984. Chick myotendinous antigen.
II. A novel extracellular glycoprotein complex consisting of large
disulfide-linked suhunits. J. Cell Biol. 98:1937-1946.
Chiquet-Ehrismann, R., E. Mackie, C. Pearson, and T. Sakakura. 1986.
Tenascin: an extracellular matrix protein involved in tissue interactions
during fetal development and oncogenesis. Cell. 47:131-149.
Crossin, K. L., S. Hoffman, M. Grumet, J.-P. Thiery, and G. M. Edelman. 1986. Site-restricted expression of cytotactin during development
of the chicken embryo. J. Cell Biol. 102:1917-1930.
Erickson, H. P., and J. L. Iglesias. 1984. A six-armed oligomer isolated
from cell surface fibronectin preparations. Nature (Land.). 311:267-269.
Erickson, H. P., and H. C. Taylor. 1986. Brachionectin: a cell attachment
protein from chicken fibroblasts and human tumors with a multi-armed
oligomeric structure. J. Cell Biol. 103(5, Pt. 2):434a. (Abstr.)
Fowler, W., and H. Erickson. 1979. Trinodular structure of fibrinogen:
confirmation by both shadowing and negative stain electron microscopy.
J. Mol. Biol. 134:241-249.
Gibson, W. 1981. Protease-facilitated transfer of high-molecular weight
proteins during electrotransfer to nitrocellulose. Anal. Biochem. 118:1-3.
Grumet, M., S. Hoffman, K. L. Crossin, and G. M. Edelman. 1985.
Cytotactin, an extracellular matrix protein of neural and non-neural tis-
16.
17.
18.
19.
20.
21.
22.
23.
24.
sues that mediates glia-neuron interaction. Proc. Natl. Acad. Sci. USA.
82:8075-8079.
Kruse, J., G. Keilhauer, A. Faissner, R. Timpl, and M. Schachner. 1985.
The J1 glycoprotein: a novel nervous system cell adhesion molecule of
the L2/HNK-1 family. Nature (Land.). 316:146-148.
LaVail, L. H., and M. W. Cowan. 1971. The development of the chick
optic tectum: I normal morphology and cyto architectonic development.
Brain Res. 28:319--419.
Laemmli, U. K. 1970. Cleavage of structural proteins during the assembly
of the head of bacteriophage T4. Nature (Land.). 227:680-685.
Maier, A., and R. Mayne. 1987. Distribution of connective tissue proteins
in chick muscle spindles as revealed by monoclonal antibodies: a unique
distribution of brachionectin. Am. J. Anat. In press.
Merril, C. R., D. Goldman, S. A. Sextman, and M. H. Ebert. 1981.
Ultrasensitive stain for proteins in polyacrylamide gels shows regional
variation in cerebrospinal fluid proteins. Science (Wash. DC). 211:
1437-1438.
Odermatt, E., J. W. Tamkun, and R. O. Hynes. 1985. Repeating modular
structure of the fibronectin gene: relationship to protein structure and
subunit variation. Proc. Natl. Acad. Sci. USA. 82:6571-6575.
Ruoslahti E., E. G. Hayman, M. Pierschbacher, and E. Engvall. 1982.
Fibronectin: purification, immunochemicalproperties, and biological activities. Methods Enzymol. 82:803-831.
Ruoslahti, E., and M. D. Pierschbacher. 1986. Arg-gly-asp: a versatile cell
recognition signal. Cell. 44:517-518.
Sanes, J. R., M. Schachner, and J. Covault. 1986. Expression of several
adhesive macromolecules (N-CAM, L1, J 1, NILE, uvomorulin, laminin,
fibronectin, and a heparin sulfate proteoglycan) in embryonic, adult, and
denervated adult skeletal muscle. J. Cell Biol. 102:420--431.
Downloaded from jcb.rupress.org on February 21, 2013
The Journal of Cell Biology, Volume 105, 1987
1394